Single‐molecule imaging reveals the translocation and DNA looping dynamics of hepatitis C virus NS3 helicase

Abstract

Non‐structural protein 3 (NS3) is an essential enzyme and a therapeutic target of hepatitis C virus (HCV). Compared to NS3‐catalyzed nucleic acids unwinding, its translation on single stranded nucleic acids have received relatively little attention. To investigate the NS3h translocation with single‐stranded nucleic acids substrates directly, we have applied a hybrid platform of single‐molecule fluorescence detection combined with optical trapping. With the aid of mechanical manipulation and fluorescence localization, we probed the translocase activity of NS3h on laterally stretched, kilobase‐size single‐stranded DNA and RNA. We observed that the translocation rate of NS3h on ssDNA at a rate of 24.4 nucleotides per second, and NS3h translocates about three time faster on ssRNA, 74 nucleotides per second. The translocation speed was minimally affected by the applied force. A subpopulation of NS3h underwent a novel translocation mode on ssDNA where the stretched DNA shortened gradually and then recovers its original length abruptly before repeating the cycle repetitively. The speed of this mode of translocation was reduced with increasing force. With corroborating data from single‐molecule fluorescence resonance energy transfer (smFRET) experiments, we proposed that NS3h can cause repetitive looping of DNA. The smFRET dwell time analysis showed similar translocation time between sole translocation mode versus repetitive looping mode, suggesting that the motor domain exhibits indistinguishable enzymatic activities between the two translocation modes. We propose a potential secondary nucleic acids binding site at NS3h which might function as an anchor point for translocation‐coupled looping.

Abbreviations

BLM

Bloom syndrome protein

DIG

digoxigenin

EMCCD

electron multiply charge coupled device

FGE

formylglycine

G4

G‐quadruplex

HCV

hepatitis C virus

HWP

half wave plate

NS3

non‐structural protein 3

NS3h

non‐structural protein3 helicase

PBS

polarizing beam splitters

PEG

polyethyleneglycol

PIFE

protein‐induced fluorescence enhancement

QPD

quadrant photodiode

SF1

superfamily 1

SF2

superfamily 2

smFRET

single‐molecule fluorescence resonance energy transfer

ssNA

single‐stranded nucleic acids

TIRF

total internal reflection fluorescence.

Introduction

Hepatitis C virus (HCV) is a positive‐sense single‐stranded RNA (ssRNA) virus,1 and infects more than 170 million people worldwide.2 Among at least 10 encoded proteins in its genome,3 nonstructural protein 3 (NS3) is an essential component of HCV viral replication complex.4 It is a bi‐functional enzyme, which contains a helicase activity in the C‐terminal and a protease activity in the N‐terminal portion. Both enzymatic functions are vital for the HCV life cycle. Recent studies and clinical trials have shown that targeting HCV NS3 is an effective anti‐viral strategy to develop promising therapeutic agents.5, 6, 7

Besides being a potential target for drug design, the helicase domain of NS3 (NS3h, 450 amino acids) serves as a widely studied prototype system of superfamily 2 (SF2) helicases which unwind duplex double‐stranded nucleic acids using the energy release during ATP hydrolysis.8 In its monomeric form, the NS3h can translocate along single‐stranded nucleic acid strands and displace other DNA‐bound proteins on nucleic acid in the 3′ to 5′ direction.9, 10 Similar to other members of the NS3/NPH‐II family, NS3 also contains a core structure of two RecA‐like α/β domains [domain 1 (D1) and 2 (D2)] with the ATPase activity at the junction of these two domains and an additional α‐helical domain [domain 3 (D3)] which can bind onto D1 and D2, embracing a nucleic acid strand.11, 12, 13, 14 A feature of NS3 that distinguishes it from other members of the family is a N‐terminus protease domain which is located at the vertex of D1, D2, and D3 domains.15, 16 In addition, NS3 is able to unwind both DNA and RNA substrates.17

As a model of SF2 helicase, the DNA/RNA unwinding activity of NS3 along with its ATP‐powered translocation has received much attention in the past decade. Snapshots of NS3 in action from X‐ray crystal structure studies of NS3 bound to single‐stranded RNA (ssRNA) or single‐stranded DNA (ssDNA) showed that NS3/NS3h interacts with nucleic acids primarily through their phosphodiester backbones.11, 18, 19 This explains the similar, although not identical, kinetics properties of NS3/NS3h on DNA and RNA.10, 17, 18 Based on structural data19, 20 and single‐molecule studies,21, 22, 23 the mechano‐chemical mechanism of NS3 appears to require one ATP‐hydrolysis per nucleotide translocation in the 3′ to 5′ direction, unwinding one base pair at a time22 or three base pairs at a time,21 depending on the conditions. Compared to the unwinding‐coupled translocation mechanism, the sole translocation process received less attention so far. It is, however, important to investigate NS3's translocation mechanism on long single‐stranded nucleic acids due to its capability to displace proteins during translocation9 as well as the important involvement of NS3 in virus particle assembly.24

Though previous studies revealed important biochemical properties such as translocation rate, step size, processivity, and the influences of sugar and base moieties of target nucleic acids on NS3h translocation on single‐stranded nucleic acid,10, 18 two major factors limited their interpretation: (1) inability to detect possible heterogeneity in translocation rate with ensemble experimental approaches which may result in misinterpretation and (2) data analyzed through kinetic models are dependent on the model assumptions. Here, we present a platform to directly observe and quantify the translocation dynamics of single NS3h molecules on thousands of nucleotides long single‐stranded nucleic acid. With the aid of optical trapping, combined with fluorescence microscopy, we were able to track single fluorophore‐labeled NS3h with high positional accuracy and compare translocation rates and processivity on ssDNA and ssRNA substrates.

Unexpectedly, a novel mode of translocation dynamic of NS3h, that we refer to as repetitive looping, has been observed on ssDNA, and single molecule FRET (fluorescence resonance energy transfer) corroborated our findings. This mode of action might share similar biological functions with the repetitive looping observed for other helicases such as PcrA,25 UvrD,26 and Pif1.27

Results

Single beam optical trapping with fluorescence sub‐pixel localization accuracy

To measure the translocation dynamics of single NS3h on long single‐stranded nucleic acids (ssNA) in real‐time, we used a single beam optical trap combined with total internal reflection fluorescence (TIRF) microscopy to detect the fluorescence of labeled molecules translocating on long ssNA as reported earlier for observing translocation and unwinding of DNA by UvrD28 as well as one‐dimensional diffusion of SSB on DNA.29 The external force of an optical trap allows stretching ssNA that would otherwise be coiled up. Two lasers (1064 nm for optical trapping and 532 nm for fluorophore excitation) are combined with a dichroic mirror and guided into a microscope objective. The force magnitude is determined by the displacement of the trapped bead from the trap center (detected by a quadrant photodiode) multiplied by the calibrated trap stiffness. The sample is excited through an objective‐type TIRF configuration and fluorescence emission is imaged using an electron multiplying charge coupled device (EMCCD) camera [Fig. 1(A)]. We fitted single fluorescence molecule spots using a 2D Gaussian function to obtain an accurate position estimation of fluorescently labeled proteins (15 nm precision, 20 ms time resolution).28 More details can be found in the Experimental Procedures.

Figure 1.

Optical layout and experimental scheme. A: Optical layout of the hybrid optical trap–TIRF‐instrument. Both the fluorescence excitation (532 nm, green) laser and the trapping laser (1,064 nm, red) were modulated by half wave plate (HWP) and polarizing beam splitters (PBS). A dichroic mirror (D1) was used to guide the two laser beams into the microscope from a side port. The scattered and non‐scattered light of the trapping laser from trapped beads is acquired at the quadrant photodiode (QPD) which reports the readout of applied forces. The desirable force is controlled by the displacement of biotin‐end ssNA tether to the trap center through an in‐house program adjusting the movement of nano‐stage. With objective‐type TIRF excitation, the fluorescence emission is back collected through objective and imaged onto an electro‐multiplying charge‐coupled device (EMCCD) camera. D1‐3, dichroic mirrors; F1‐F2, filters; T1, telescope. B: 5′‐end single‐stranded DNA is anchored onto polyethylene glycol (PEG) passivated coverglass surface through biotin‐neutravidin linkage and 3′‐tail of single‐stranded DNA is coupled to polystyrene bead by antigen–antibody interaction, digoxigenin (DIG), and anti‐digoxigenin (Anti‐DIG). Fluorescence excitation laser (green) excites the coverglass within a few hundred nanometers by TIR. During measurements, trapping laser (red) pulls out the linked bead and stretches single‐stranded nucleic acid linearly. NS3h is labeled with Cy3 dye as fluorescence tracking marker.

Direct visualization of NS3h translocation on long single‐stranded nucleic acids

We directly visualized the translocation of single NS3h (a truncated version of NS3 which comprises the entire helicase domain) molecules along long single‐stranded nucleic acids with our hybrid assay [for experimental scheme, see Fig. 1(B)]. The long single‐stranded nucleic acid is bound to an anti‐digoxigenin coated polystyrene bead through antibody–antigen interaction. The 5′ end of the DNA (3′ end for RNA) is labeled with biotin and anchored to the surface via biotin–neutravidin linkage. Different strand polarities are used between ssDNA and ssRNA experiments due to the differences in the synthesis approaches: the polarity from the surface to the trapped bead is 5′ to 3′ for ssDNA and 3′ to 5′ for ssRNA. The long single‐stranded nucleic acids which are stretched by the optical trap serve as one‐dimensional tracks for the locomotion of NS3h [Fig. 2(A)]. The detection of the moving helicases is based on fluorescence labeling of NS3h. The angle θ between the single‐stranded nucleic acid and the surface can be calculated as the arctangent of the trapped bead height (660 nm) and the horizontal distance from the trapped bead center to the surface anchoring point of the ssNA which ranged from 0.063 to 0.126 radians. All projected enzyme movements acquired are corrected by the conversion factor of arccosineθ, which typically lies between 3 and 7 degrees.

Figure 2.

NS3h translocation visualization on long ssDNA. A: NS3h translocation. NS3h‐Cy3 translocates along long ssDNA from 3′ to 5′ end. B: Kymogram showing multiple NS3h translocation events. The color map represents the fluorescence intensity of tracked molecules. NS3h translocates in 3′ (top) to 5′ (bottom) direction. Green lines are fitting results of a 2D Gaussian function showing the center‐of‐mass of the moving molecules. C: Representative traces of NS3h translocation. Fitting fluorescence traces with a 2D Gaussian function allows to probe NS3h translocase activity at sub‐pixel accuracy. Slopes of black lines obtained by linear fitting represent the translocation speed of NS3h. The range of distribution of slopes indicates the molecular heterogeneity of monomeric NS3h.

In order to measure helicase movement, a single Cy3 fluorophore is attached to the C‐terminus of NS3h using the aldehyde‐tag method30, 31, 32, 33 and 2 nM fluorophore‐labeled NS3h is added to the reaction chamber together with 2 mM ATP. During experimental recording, fluorophores appear as Gaussian‐fittable spots recorded by an EMCCD camera and move along a linear track. Figure 2(B) is a typical kymograph, which shows the time course of the movement of fluorescently labeled NS3h. The kymograph demonstrates that a NS3h monomer translocates toward the ssDNA's surface tethering point away from the trapped bead (both marked) in the presence of ATP, consistent with 3′ to 5′ translocation.9 ATP was required for observation of translocation events. The gradual increase of fluorescence intensity when NS3h approached the surface is due to the TIR configuration with the excitation intensity decreasing exponentially away from the surface.28 We discarded NS3h translocation events showing more than one photobleaching step in the fluorescence trace in order to reduce the possibility that the analyzed trajectories are due to NS3h oligomers.

In order to get the precise position and intensity trajectories of NS3h molecules, we applied two‐dimensional (2D) Gaussian fitting. Figure 2(B) presents an example of fitting a 2D Gaussian to a fluorescence image at each time point and overlaying the longitudinal position determined on a kymograph (green line). For each translocation event, linear fitting is applied to the fitted longitudinal position vs. time curve. The individual translocation rates can be derived from each fitting [several representative translocation trajectory traces on ssDNA are shown in Fig. 2(C)]. In order to convert the translocation rate from nanometers per second (nm/s) to nucleotides per second (nt/s), the length of long single‐stranded nucleic acids, which varies between molecules, needs to be determined. We used the worm‐like chain model to fit the extension vs. force curve of each molecule to obtain an estimate for the length and the conversion factor, in nt/nm, as a function of tension. Multiplying the conversion factor with the fitting results of translocation events, we calculated the average translocation rate of each NS3h molecule. Figure 3(A) shows the distribution of translocation rates in nt/s vs. applied tension obtained for ssDNA and ssRNA in the presence of 2 mM ATP. In the range from 12 to 20 pN, the translocation rate is nearly constant, suggesting that the rate‐limiting step of NS3h translocation is not tension‐dependent in this range. We observed a 20% decrease at 8 pN compared to the average from 12 to 20 pN, but presently we cannot rule out the possibility that a small change such as this may have occurred due to inaccuracies in the determination of the DNA or RNA length from worm‐like chain model fitting. Therefore, we calculated the average translocation rates by combining data obtained at all force values examined, where the mean translocation rate was 24.4 ± 0.8 nt/s on ssDNA and 74.0 ± 0.7 nt/s on ssRNA [Fig. 3(B)]. The ssDNA translocation rate we observed is similar to previously published results (46 nt/s on ssDNA)10 and is consistent with the tendency of NS3h to translocate more rapidly on ssRNA.18 The processivity of NS3h translocation could be inferred from the histogram plot of NS3h translocation distance which is determined as the distance covered from binding of NS3h, which is detected as a sudden appearance of a fluorescence spot, until the disappearance of fluorescence, either caused by NS3h dissociation or photobleaching. By plotting the probability of fluorescent spots remaining vs. distance traversed, we were able to calculate a lower bound for the processivity of NS3h [324 ± 9 nt on ssDNA, 488 ± 13 nt on ssRNA; Fig. 3(C)]. Because of the three times faster translocation on ssRNA, the average duration of translocation events is shorter on ssRNA by about a factor of two.

Figure 3.

Statistics of NS3h translocation events on long ssNA. A: Translocation rate measured as a function of applied tension to ssNA. The mean translocation rate and s.e.m. are listed sequentially and respectively, ssDNA: 21.0 ± 2.5 nt/s, 25.5 ± 2.1 nt/s, 26.9 ± 1.2 nt/s, and 27.1 ± 2.9 nt/s; ssRNA: 69.4 ± 6.1 nt/s, 77.2 ± 2.7 nt/s, 78.9 ± 2.8 nt/s, and 78.5 ± 4.3 nt/s. The conversion factors of nm/s to nt/s as a function of tensions were obtained by worm‐like chain modeling. B: Distribution of NS3h translocation rate observed at difference forces. Gaussian functions (light gray and black curve) were fitted to the histograms. Mean and s.d. of ssDNA: 24.3 nt/s and 0.8 nt/s, respectively; mean and s.d. of ssRNA: 74.0 nt/s and 0.7 nt/s, respectively. C: Processivity of NS3h translocation. The histogram shows the fraction of remained molecule after translocating a given distance. The translocation processivity of NS3h was determined by the fitting of exponential decay (light gray and black curve), an inverse of decay constant, 324 ± 9 nt (ssDNA) and 488 ± 13 nt (ssRNA).

Novel mode of NS3h translocation on ssDNA

When the optical trap stretches long ssDNA, a feedback system is used to drive the piezo nanopositioning stage to adjust the distance between the surface‐tethering point of the DNA and the center of the trapped bead in order to maintain a constant force [Fig. 4(A)]. During the recording of translocation events, we observed that a subpopulation of ssDNA undergoes repetitive DNA shortening [Fig. 4(B)] where the DNA extension gradually decreases before an abrupt increase to the original length and this process repeats itself multiple times (raw data and fitting results are shown in gray and orange, respectively). When the DNA extension snaps back to its original length, a force jump and its subsequent relaxation is observed [Fig. 4(B), raw data in gray and smoothed curve in black] due to the finite time it takes for the feedback system to respond. ATP was required for the observation of repetitive DNA shortening. The average shortening distance per cycle was 150 ± 6 nt [Fig. 4(C)] which is much shorter than the average distance traversed in ssDNA translocation events. The shortening rate decreased monotonically with force, dropping by about 40% when the force increases from 8 pN to 20 pN [Fig. 4(C)], suggesting the opposing force slows down the rate limiting steps in DNA shortening. We did not observe RNA shortening events in experiments performed using ssRNA.

Figure 4.

NS3h‐mediated ssDNA shortening. A: Schematic representation of NS3h‐mediated ssDNA shortening. Long ssDNA was stretched with the optical trap at constant force. To keep the force constant, feedback control allowed to adjust the distance between trap center and the 5′‐end of ssDNA automatically. Stage position recording allowed the observation of the shortening effect. B: Force trajectory and time‐lapse recording of ssDNA length. The ssDNA tether length gradually decreases and suddenly increases in a periodic pattern. The detected force keeps in a constant while tether length of ssDNA gradually decreases, and force relaxation occurs when the total tether length of ssDNA increases. The shortening of ssDNA length is mediated by NS3h. The increase in tether length shown by force relaxation indicates that NS3h either fully or partially dissociates from the ssDNA. C: Processivity of NS3h‐mediated ssDNA shortening. The shortened distance mediated by NS3h was determined by fitting first order decay function (light gray curve), the inverse of the decay constant is 150 ± 6 nt. D: Shortening rate measured as a function of applied force to ssDNA. Mean shortening rate and s.e.m. are listed: 39.1 ± 1.9 nt/s, 32.7 ± 2.5 nt/s, 24.6 ± 0.6 nt/s, 23.7 ± 1.1 nt/s. Force‐dependent behavior can be seen.

Mechanism of NS3h‐mediated repetitive‐looping on ssDNA

In order to examine the mechanism of the DNA shortening phenomenon, we designed a partial duplex DNA construct to perform a single‐molecule FRET (smFRET) assay [Fig. 5(C)]. A DNA construct of a long‐ and a short strand which, hence, form a partial duplex was immobilized on a PEG‐passivated surface via biotin–neutravidin interaction. The donor fluorophore, Cy3, is located at the 5′ end of the long strand and the acceptor fluorophore, Cy5, is attached at the 3′ end of the short strand. The FRET efficiency, E _FRET, reports the time‐averaged distance between the two fluorophores [Fig. 5(A)]. For a single‐stranded 5′ tail of 65 nt, d(T)_65, an E _FRET of ∼0.12 is observed in the absence of NS3h. When NS3h and ATP are injected into the reaction chamber, two distinct subpopulations are observed. Subpopulation 1 [Fig. 5(A)] showed a gradual increase of Cy3 fluorescence intensity followed by a rapid drop, and this asymmetric pattern repeated itself multiple times without observable pauses in between. The gradual increase in Cy3 fluorescence intensity is likely caused by a protein that is approaching Cy3 at the 5′ end of the ssDNA tail over time. This effect is known as protein‐induced fluorescence enhancement (PIFE), resulting from protein‐induced steric constraints, suppressing a nonradiative decay pathway thereby increasing the emission quantum yield of Cy3.34, 35 The quick drop in Cy3 intensity indicates that once NS3h reaches the end of the 5′ ssDNA tail, the protein snaps back to the initial position before the cycle is repeated [Fig. 5(C)]. This behavior, which we term “repetitive translocation,” continues until NS3h dissociates or Cy3 is photobleached, reminiscent of repetitive shuttling of Escherichia coli Rep helicase observed on a 3′ ssDNA tail.36 Direct excitation of Cy5 confirmed that Cy5 is fluorescently active. Subpopulation 2 [Fig. 5(B)] presented a similar pattern as Subpopulation 1, except that, instead of Cy3, here, E _FRET gradually increased and was followed by a quick drop and repetition. During E _FRET change, we also observed that the sum of both donor and acceptor increased, likely due to PIFE. The mode of repetitive cycling indicates that two ends of ssDNA, which are Cy3 and Cy5‐labeled, are gradually brought close to each other and then released periodically. Figure 5(D) shows a reel‐in model, which is similarly observed for PcrA, a 3′ to 5′ helicase from Bacillus stearothemophilus.25 N3Sh is statically positioned at the duplex junction and begins to reel‐in the 5′ tail using its motor domains. This results in Cy3 getting closer to Cy5, which is located at the duplex junction, causing a gradual increase in E _FRET. Moreover, the behavior of Subpopulation 2 appears to be related to the repetitive DNA shortening phenomenon detected in the optical trapping measurements under constant force because such ssDNA reeling‐in events would give rise to a gradual shortening of the distance between the trapped bead and the surface‐tethering point. Therefore, we propose that the repetitive shortening events in the optical trapping experiments are caused by the formation of ssDNA loops induced by NS3h ssDNA translocation. The distribution of the ssDNA shortening length can be referred as the loop size formed by NS3h [Fig. 4(C)]. The histograms of time intervals of each repetition cycle for both subpopulations shows that their average cycle duration is similar, 2.15 ± 0.03 s (Subpopulation 1) and 1.7 ± 0.05 s (Subpopulation 2) [Fig. 5(E,F)].

Figure 5.

NS3h‐mediated repetitive‐looping. A: Cy3 intensity increases as NS3h approaches. Two populations were observed in single‐molecule time traces in smFRET experiments. Subpopulation 1: the fluorescence intensity of Cy3 fluorophore shows a gradual increase followed by a quick drop repeatedly which is caused by the PIFE effect. smFRET time traces were recorded in the presence of 2 mM ATP and 2 nM NS3h. Donor (Cy3) intensity is shown in green, acceptor (Cy5) intensity in red, and FRET efficiency in blue. B: FRET values show the distance changes between the donor fluorophore at the 5′‐tail and the acceptor fluorophore at the partial duplex junction. Subpopulation 2 shows repetitive cycling in single‐molecule time traces, in addition. C: Repetitive translocation model. (II) NS3h binds at the 5′ partial duplex junction and starts to translocate along the ssDNA, d(T)65, until it reaches the 5′end (Cy3). Once NS3h reaches the end, it snaps back to the duplex junction and re‐initiates translocation. D: Reeling‐in model. (I) NS3h binds to the partial duplex junction. (II) Translocation starts in the 3′ to 5′ direction while NS3h stays at the duplex junction. (III) The size of the ssDNA loop formed increases while NS3h continues to translocate to the 5′‐tail. (IV) NS3h reaches the end of the nucleic acid and dissociates. (I)–(IV) are repeated for numerous cycles. E: Histogram of time intervals of the repetitions (Δt) shown in (A) and fitted using a Gaussian function (black curve; mean time interval: 2.15 ± 0.03 s). F: Histogram of time intervals of the repetitions (Δt) shown in (B) and fitted using a Gaussian function (black curve; mean time interval: 1.70 ± 0.05 s).

Discussion

Difference of translocation rates between ssDNA and ssRNA

We have characterized the translocation dynamics of NS3h on long ssNAs, including binding duration, translocation rate, and processivity via direct fluorescence imaging of labeled NS3h. In addition, we identified repetitive‐looping as a yet unreported mode of translocation of NS3h. The imaging analysis showed that NS3h monomer translocates at a rate of 24.3 ± 0.8 nt/s on ssDNA and 74.0 ± 0.7 nt/s on ssRNA [Fig. 3(B)], broadly consistent with the previous estimates obtained using ensemble experimental approaches. Matlock et al. used different lengths of fluorescence labeled poly(dT) to estimate an NS3h translocation rate of 46 ± 0.7 nt/s.10 Khaki et al. applied a similar approach on poly(dT), poly‐deoxyuridylate [poly(dU)], and poly‐uridylate [poly(U)] and estimated the translocation rates of 3.35 ± 0.09, 35.4 ± 0.6 and 42.2 ± 1.5 nt/s, respectively.18 The variations might be in part explained by differences in the ionic conditions and the experimental temperatures [100 mM in our case versus 50 mM ¹⁰ versus 30 mM ¹⁸ (NaCl); 22°C versus 37°C¹⁰ versus 25°C¹⁸].

Consistent with a previous report,18 we found that the rate of nucleic acid translocation by NS3h is dependent on the sugar moiety of the nucleic acid: the translocation rate for NS3h is approximately three times higher on ssRNA than on ssDNA. However, NS3 interacts with ssNAs mainly through their phosphate backbone. How do we then explain the difference in translocation rates between ssRNA and ssDNA? We compared structures of NS3h–ssDNA (PDB ID: 3KQH)20 and NS3–ssRNA (PDB ID: 3O8C)19 in order to get insight into why the NS3 translocation rate on ssNAs is dependent on the sugar moiety. Although the two crystal structures differ in the inclusion of the protease domain in the NS3–ssRNA structure only, the protease domain increases RNA binding affinity but has no direct effect on ATPase activity.37 Domain D3 of NS3 (residues 487–623) shares a high structural similarity in the structures of NS3h in complex with ssDNA compared to NS3 in complex with ssRNA.19 We therefore structurally aligned the two structures with D3 (residues 487–623) as a frame of reference using MultiSeq.38 The overlay is colored based on the structural identity Q per residue (Qres)39 obtained in the alignment [Fig. 6(A)]. In contrast to the well aligned Domain 2, Domain 1 in NS3h–ssDNA shows a 2–5 Å deviation from Domain 1 in NS3–ssRNA. The most dramatic structural changes are located at motif V (ssNA binding, Domain 2) and motif I (ATPase site, Domain 1).

Figure 6.

Single‐stranded nucleic acid‐dependent conformational changes of NS3h. A: Structural alignment of NS3‐ssRNA and NS3h‐ssDNA (PDB ID: 3O8C, 3KQH, respectively). Carbons of ssRNA and ssDNA are colored in cyan and orange accordingly, other atoms are colored by atom type (dark blue, nitrogen; red, oxygen; tan, phosphate). Residues 487 to 623 of domain 3 (D3) were used for the structural alignment. The scale bar shows the calculated Qres values, which indicate structural similarity. B: Local conformational and interaction changes of the single‐stranded nucleic acid in complex with NS3/NS3h. The sidechain and backbone atoms of NS3/NS3h residues that interact with nucleotides are shown with sticks and are labeled with different colors (black dashed lines refer to the same interacting residues between NS3–ssRNA and NS3h–ssDNA; cyan dashed lines refer to uniquely interacting residues in the NS3–ssRNA structure; orange dashed lines show interactions in the NS3h–ssDNA structure). Motif V (residues 410 to 419) of NS3–ssRNA (cyan) and NS3h–ssDNA (orange) are shown as cartoon coils. Nucleotide residues (NA1 to NA5) and the 5′ and 3′ ends of the strands are labeled. C: Surface charge distribution of NS3h–ssDNA complex. The accessible surface of NS3h–ssDNA is colored based on electrostatic potentials (positive charge, blue; negative charge, red; neutral, white). The molecular view with D1 and D2 on the bottom, and the D3 on the top is similar to that in (A). The reverse side of the NS3h–ssDNA complex is shown in (D). The electrostatic surface was calculated using the program VMD with APBS plugin.

Six interaction sites are conserved between the two structures19 [Fig. 6(B), black, dashed lines]. Except for the pi‐stacking interaction of Trp501 with the nucleic acid, the remaining interactions of conserved amino acids occur via hydrogen bonding (Val432, Thr411, Arg393, Val232, and Gly255). A significant change in the local secondary structure of the protein is observed in motif V [Fig. 6(B)]. In the structure of NS3–ssRNA, motif V forms a compact helical structure while in NS3h–ssDNA, motif V forms an extended loop segment. Motif V interacts with the ssNAs through hydrogen bonds at Thr416 (cyan, dashed line) and Leu414 (orange, dashed line) with RNA and DNA, respectively. The more compact structure in motif V of ssRNA nucleic acids–protein complexes may account for the lower entropic contribution compared to ssDNA complexes as measured by isothermal titration calorimetry.18

The conformational difference between deoxyribose and ribose can be observed by overlaying the ssNA–protein complexes.19 For ssDNA, nucleic acid 1 (NA1) and NA2 adopt a C2′‐endo pucker conformation while the rest of the ssDNA (NA3, NA4, NA5) are in a C3′‐endo pucker conformation. In contrast, the entire ssRNA in an NS3–ssRNA complex maintains a uniform C3′‐endo pucker conformation. The C2′endo conformation produces a looser backbone than the C3′endo conformation which results in a longer interphosphate distance between adjacent bases.40 Another interesting observation is the extra intramolecular hydrogen bonds found within ssRNA, between the 2′‐hydroxy groups and the 4′‐oxygen atoms of ribose moieties. Together, the formation of extra intramolecular hydrogen‐bonds and the C3′‐endo conformation of the ribose rings results in a more compact structure of the backbone of ssRNA compared to ssDNA.

To summarize, the differences of nucleic acid binding affinities of NS3h to nucleic acids between ssDNA and ssRNA can be well explained by the protein structure, namely the ssNA–protein interaction sites, and the conformation of sugar moieties. Khaki et al.18 noted that the affinity of nucleic acid binding is inversely correlated with the NS3/NS3h translocation rate. Moreover, their nucleic acid binding data showed that NS3h–ssDNA complex has stronger affinity than NS3h–ssRNA complex. Hence, the differences of the translocation rates between deoxyribose and ribose in our experiments may arise from the more compact structure of ssRNA which results in weaker binding compared to ssDNA. The low binding affinity of NS3h–ssRNA may also be the reason that we did not observe repetitive‐looping subpopulation with RNA substrate. Without a stable binding site, the formation and maintenance of nucleic acid loop can be difficult.

In both ssDNA and ssRNA, the lower bound for the average translocation distance determined from our single molecule imaging was higher than estimations from ensemble experiments [324 ± 9 nt versus 230 ± 60 nt(DNA)10; 488 ± 13 nt versus 150 ± 30 nt (RNA)18]. An assumption in the previously reported kinetic ensemble analyses is that there is only one translocation mode. As we discuss in detail below, we observed a new mode of translocation of NS3h on ssDNA that we termed “repetitive‐looping,” which may in part account for the discrepancies in ssDNA translocation distance between our direct measurements and estimates from bulk experiments.

Possible mechanism of repetitive‐looping of NS3h and its consequences

Unexpectedly, we discovered that adding NS3h to a long stretched ssDNA can result in progressive shortening of DNA followed by an abrupt recovery, and this cycle repeated multiple times. A single molecule FRET experiment using a 5′ (dT)₆₅ tailed DNA showed repetitive looping dynamics, corroborating the optical trap data, an NS3h‐mediated ssDNA shortening effect. A higher concentration of NS3h is used in optical trapping experiments than in smFRET experiments, due to the imperfect labeling efficiency of NS3h. Therefore, we cannot rule out the possibility that the dimerization41 or even oligomerization42 of NS3h is the cause of ssDNA shortening effect. The ssDNA used lacks significant secondary structure elements such that hairpin formation is unlikely. The average distance traversed during looping was much shorter than in ssDNA translocation (150 ± 6 nt versus 324 ± 9 nt). Moreover, a clear force‐dependent looping rate is observed when different tensions are applied to the ssDNA [Fig. 4(D)].

In smFRET experiments, due to the configuration of the partial duplex and the 5′‐overhang of the ssDNA construct, a similar reel‐in model like for PcrA25 is proposed here [Fig. 5(D)]. PcrA is a well‐studied superfamily 1 (SF1) helicase and known for its inchworm‐like translocation mechanism and looping mode.25 Although the exact domain movements and structural details are different, we believe that NS3h shares a very similar mechanism for looping translocation. Dwell‐time analysis from Subpopulation 1 (translocation) and Subpopulation 2 (repetitive‐looping) yielded the average cycle times of 2.15 ± 0.03 s and 1.70 ± 0.05 s on 65 nt, respectively, corresponding to the movement speed of 30.2 nt/s (translocation) and 38.2 nt/s (repetitive‐looping). The rate derived from the smFRET experiment for Subpopulation 2 agree well with the rate derived from the optical trapping experiment of repetitive looping, suggesting that they share a common mechanism. We hypothesize, based on the surface charge distribution of NS3h [Fig. 6(C,D)], that a potential second nucleic acid binding site on D3 might explain loop formation. In this scenario, the backbone of the ssNA may bind at a certain position on the secondary binding site via electrostatic interaction while motor domains (D1 and D2) continuously hydrolyzes ATP to promote the movement of the reeled‐in nucleic acid. Then, once the ssNA dissociates from the hypothetical secondary binding site on D3, the looping event reaches an end. It is also possible that repetitive‐looping occurs without a secondary binding site, which has been speculated in an early study.15 If ssDNA occasionally remained bound to the surface of one of RecA‐like domain while the other RecA‐like domain moved forward through many cycles of ATP dependent conformational changes, a ssDNA loop can develop.

The repetitive movements of helicases along the same segment of DNA or RNA have been observed from different helicase superfamilies and various organisms. Examples include Rep,36 UvrD,26 and PcrA25 which translocate on ssDNA, Pif1 which translocates on ssDNA and unwinds G‐quadruplex structures27 and RIG‐I, the human viral RNA sensor, that translocates on dsRNA without unwinding it.43 It appears that all helicases listed in Table 1 are able to move or translocate repetitively on nucleic acid substrates without full dissociation. We suspect that the biological function of NS3 repetitive‐looping is the removal of bound proteins as was proposed for Rep and PcrA but further investigation is needed to test this proposal.

Table 1.

Helicase Showing Repetitive Translocation/Unwinding

Helicase	Origin	Repetition observed	Repetition mechanism	Functional implication
NS3	Viral	dsDNA unwinding/ssDNA translocation	Holding on to unwound strand/Duplex‐anchored looping and release	Unknown
UvrD	Bacterial	ssDNA translocation	Duplex‐anchored looping and release	Removal of RecA filament
Rep	Bacterial	ssDNA translocation	Blockade‐induced 3’ capture	Prevention of RecA filament formation
PcrA	Bacterial	ssDNA translocation	Duplex‐anchored looping and release	Removal of RecA filament
Pif1	Yeast	G4 unwinding	Duplex‐anchored looping and release	Keep G4 unfolded during replication
RIG‐I	Human	dsRNA translocation	Unknown	Viral RNA recognition and signaling
BLM	Human	dsDNA unwinding	Strand switching	Unknown

Materials and Methods

Instrument details

The main instrument used in this study is a single beam optical trap combined with an objective‐type total internal reflection (TIR) fluorescence microscope.28 A 1064 nm Nd‐YAG trapping laser was used to stretch ssNAs attached to a PEG‐passivated cover slide. A 532 nm diode‐pumped solid state laser was used for fluorescence excitation. The reaction chamber was observed through an oil‐immersion objective lens (×100/1.4, Olympus). An XYZ nano‐stage (Mad City Labs) was mounted on the microscope stage and controlled by an in‐house program to apply mechanical force to tethered nucleic acids bound to polystyrene beads. The scattered and non‐scattered laser light of trapped bead were collected by a condenser and imaged onto a two dimensional QPD (Pacific Sensors) to provide the readout of forces. The trap stiffness was determined using the power spectrum method.44 The emission of fluorescence of labeled molecules (NS3h‐Cy3) was recorded by an EMCCD camera (Andor iXon) at a frame rate of 50 Hz. The TIR configuration provided a formidable signal to noise ratio, thus allowing to fit two‐dimensional (2D) Gaussian functions onto the fluorescence spots.45 Through this approach, sub‐pixel precision in translocation measurements of labeled NS3h could be achieved.

Long single‐stranded nucleic acid synthesis and purification

The ssDNA substrate was synthesized by rolling circle amplification.46 A circular ssDNA template was used and a 5′‐end biotinylated primer was hybridized to the template. phi29 DNA polymerase (NEB) was used to generate single‐stranded DNA in the range of tens of thousands of nucleotides. The sequence of the template and the primer were 5′‐AT ATT TT(T)7 A(T)8 TTA TTA‐3′ and 5′‐AAT AAT TAT AAA‐3′. After applying gel electrophoresis to the PCR products, purified ssDNA was incubated with digoxigenin‐11‐dideoxy‐UTP (Roche Applied Science) in order to label it at the 3′ end.

Several kilobases‐long ssRNA (poly‐uridine, poly‐U) was synthesized from 20 mM UDP employing the reverse reaction of polynucleotide phosphorylase (Sigma‐Aldrich) at pH 9.0 (100 mM Tris‐HCL) in the presence of 5 mM MgCl₂, 0.4 mM EDTA, 0.04% wt/vol BSA, and 5′‐digoxigeninylated U₂₀‐primer at 37°C for 15 min. The reaction product was purified with an RNeasy kit (Qiagen) and 20 μg per batch of the ssRNA (5000–10,000 bases) was cleaved at the 3′‐ribose with sodium periodate chemistry under exclusion of ambient light and after quenching the oxidation with 10% glycerol conjugated with biotin‐hydrazide in the presence of RNasin and sodium cyanoborohydride at room temperature. The end product was purified via agarose gel electrophoresis. The desired fraction of the agarose gel was then extracted with a gel extraction kit (Qiagen) and used for single‐molecule studies.

Fluorescence labeling (Cy3) of NS3h and purification

The aldehyde‐tag approach31 has been employed, here, in order to label NS3h fluorescently. A formylglycine (FGE) modification was engineered at the C‐terminus of NS3h with a His6‐Smt3 modification at the N‐terminus. NS3h was sequentially purified via 6xHis‐tag affinity, ion‐exchange and gel‐filtration columns. His6‐Smt3 was cleaved off NS3h using the Ulp1 protease. For Cy3‐labeling, FGE‐modified NS3h was concentrated and directly added to Cy3 hydrazide, followed by gentle agitation, incubation at 4°C for 42 h and subsequent removal of excess dye by repeated buffer exchanges. The labeling efficiency of NS3h is between 16% to 50% in this study.

Sample chamber preparation

The sample chambers were assembled with a coverslip and a slide, taped together with double‐sided adhesive tape as spacers. The coverslips were passivated with PEG and biotin‐PEG.47 Neutravidin was injected into the reaction chambers followed by addition of 5′‐biotinylated ssDNA or 3′‐biotinylated ssRNA. Free ssDNA/RNA was washed out after 10 min incubation. Subsequently, anti‐digoxigenin coated polystyrene beads were added to capture the 3′ ends of ssDNAs or 5′ ends of ssRNAs, which were labeled with dioxigenin.

Experimental conditions

All NS3h experiments were performed in 25 mM HEPES (pH 7.7), 100 mM NaCl, 1mM DTT, and 5 mM MgCl₂ at 22°C. The concentration of NS3h was 12–20 nM for optical trapping experiments, and 2 nM for smFRET experiments. The ATP concentration was 2 mM. To increase the fluorescence lifetime and photo‐stability of Cy3 an oxygen‐scavenging system was used. [4 mM Trolox (Sigma‐Aldrich), 0.8% dextrose with glucose oxidase (Sigma‐Aldrich), catalase (Calbiochem)].

Polymer modeling

In order to estimate the conversion factor, nm to nt, for the ssNAs, we used a extensible worm‐like chain model48 to fit the ssDNA and ssRNA extension curves. Modeling was done using a persistence length of 1.00 nm (ssDNA) or 0.91 nm (ssRNA), a stretch modulus of 1000 pN (ssDNA) or 1593 pN (ssRNA) and a contour length of one nucleotide covering 0.6 nm (ssDNA) or 0.59 nm (ssRNA).49, 50